Multi-User Distributed Spectrum Access Method for 802.11ax Stations

The 802.11ax MAC incorporates Multi-User (MU) Orthogonal Frequency Division Multiple Access (OFDMA) based uplink communication where stations access spectrum by randomly selecting one of the available sub-channel, called Resource Unit (RU). This paper proposes a distributed RU selection method using Convolution Neural Network (CNN) based Deep Reinforcement Learning (C-DRL). The proposed method works in tandem with the standard CSMA/CA protocol where CSMA/CA determines when a transmission opportunity is available to a station while the proposed method is used to select the RU for transmitting the data. Specifically, each station locally trains its CNN in an online manner solely based on energy detection and acknowledgment packets. The proposed method achieves 81:18% higher average throughput and 42:37% lower latency compared to standard MU-OFDMA MAC protocol for a single Access Point network with a variable number of stations.